Should the biofilm mode of life be taken into consideration for microbial biocontrol agents?

Summary Almost one‐third of crop yields are lost every year due to microbial alterations and diseases. The main control strategy to limit these losses is the use of an array of chemicals active against spoilage and unwanted pathogenic microorganisms. Their massive use has led to extensive environmental pollution, human poisoning and a variety of diseases. An emerging alternative to this chemical approach is the use of microbial biocontrol agents. Biopesticides have been used with success in several fields, but a better understanding of their mode of action is necessary to better control their activity and increase their use. Very few studies have considered that biofilms are the preferred mode of life of microorganisms in the target agricultural biotopes. Increasing evidence shows that the spatial organization of microbial communities on crop surfaces may drive important bioprotection mechanisms. The aim of this review is to summarize the evidence of biofilm formation by biocontrol agents on crops and discuss how this surface‐associated mode of life may influence their biology and interactions with other microorganisms and the host and, finally, their overall beneficial activity.


Introduction
Approximately 30% of crop yields are lost every year worldwide, mostly due to diseases caused by pests, weeds or pathogenic microorganisms (Teng and Krupa, 1980;Teng, 1987;Oerke, 1999Oerke, , 2006Savary et al., 2012). The microbiological control of agricultural products along the food chain is still mainly ensured by the extensive use of chemical pesticides, preservatives and synthetic drugs (Horrigan et al., 2002). Environmental pollution and associated human diseases caused by this excessive use of chemicals during last century has led many agencies and governments worldwide to support an alternative route, where agriculture can be productive and economically viable, while still addressing societal and environmental concerns (Anonymous, 1999;Hazell and Wood, 2008;Aktar et al., 2009). Biological protection strategies are used and encouraged from farm to forks to prevent pathogen contaminations and livestock or crop diseases (Pal and McSpadden Gardener, 2006;Sundh and Melin, 2010;Jordan et al., 2014). Biological control, or 'biocontrol', consists in the removal of the harmful activity of one organism via one or more organisms or natural products extracted from microorganisms, plants, animals or minerals (Pal and McSpadden Gardener, 2006).
The relationship between survival, persistence and virulence of pathogenic microorganisms with their biofilm mode of life have been clearly established since the early 1980s (Costerton et al., 1978;Lam et al., 1980). According to the National Institute of Health, 80% of human infections involves microbial biofilms (NIH, 2002). Biofilmassociated infections have also been reported in agricultural settings, e.g., in crops and animal diseases (Davey and O'toole, 2000;Prigent-Combaret et al., 2012;Li et al., 2015). Indeed, the sessile mode is the preferential lifestyle of microorganisms, regardless of their biotope (Davey and O'toole, 2000;Morris and Monier, 2003). A biofilm can be described as a spatially structured community of microorganisms, generally embedded in an extracellular matrix, and adhering to a living or inert surface (Costerton et al., 1999;O'Toole et al., 2000). Biofilm formation is generally favoured in harsh environmental conditions, such as low nutritive or toxic media (Rendueles and Ghigo, 2015) and most bacteria can form biofilms in various environments (Morris and Monier, 2003;Aparna and Yadav, 2008). Staphylococcus aureus and Pseudomonas aeruginosa are two opportunistic pathogenic bacteria that cause a diverse set of diseases and are the most highly used model bacteria for biofilm studies. They can colonize the human nasopharynx and form biofilms when specific environmental conditions are met, causing invasive diseases, such as chronic pneumonia. These infections are difficult to treat because of the persistence of biofilms and their high resistance to antimicrobials (Blanchette and Orihuela, 2012;Ding et al., 2016a). Bacteria can colonize and form biofilms on stems, leaves and the rhizosphere of plants, as well as soil particles, mushrooms or organic compost (Figs 1A and 2) (Ramey et al., 2004;Prigent-Combaret et al., 2012). For example, Dickeya dadantii, the causal agent of soft rot disease in a wide range of plant species, can colonize and form biofilms on chicory leaves, causing disease due to the production of degradative enzymes (Prigent-Combaret et al., 2012;Pandin et al., 2016).
Although less explored, the formation of biofilms by moulds, yeast and algae, alone or in combination, in a variety of biotopes has also been reported (Morris and Monier, 2003;Aparna and Yadav, 2008;Zarnowski et al., 2014;He et al., 2016;Rajendran and Hu, 2016;Sheppard and Howell, 2016). Aspergillus fumigatus, a human pathogen, is a filamentous fungus that can form structured biofilms. The cohesive cement of the fungal biostructure is a polymeric extracellular matrix that protects the hyphae from the host immune system, similar to bacterial biofilms (Breitenbach et al., 2016;Mitchell et al., 2016;Sheppard and Howell, 2016;Shirazi et al., 2016). Fusarium oxysporum f. sp. cucumerinum, the pathogen responsible for cucumber Fusarium wilt, can also grow inter-and intracellularly, allowing the rapid colonization of the plant and biofilm formation . Until recently, efforts in biofilm research have The first step involves deposition on the substratum (1) followed by adhesion (2) to the support through cell wall decorations and extracellular appendages. Once attached, a proliferation phase (3) and the diversification of cell types initiate the spatial organization of the biostructure, leading to biofilm maturation (4). Biofilm ageing or environmental conditions unfavourable for the maintenance of the biofilm results in regulated dispersion of the biofilm (5), disseminating free cells and cell clusters that will start a new biofilm cycle on a new surface. B-D. focused mainly on the medical field and essentially towards their eradication. With the emergence of biocontrol in agriculture, many microbiological products have been developed and are used in fields (Borriss, 2015). The main way of action of most of these commercial products is the antagonistic effect of antimicrobial molecules secreted by the biocontrol agent (Chowdhury et al., 2015;Mora et al., 2015). However, recent research in this field has made it possible to consider other major biological processes, including biofilm formation of biocontrol agents in crops (Bais et al., 2004;Bogino et al., 2013;De la Fuente et al., 2013).

Evidence of biofilm formation on crops by biocontrol agents
There is ongoing research to identify new biocontrol agents from environmental isolates and numerous biocontrol products have been developed and put on the agricultural market, mostly in Europe and North America (Borriss, 2015). Various products are in use and are effective on a wide range of plants. These include biofungicides, bactericides and biofertilizers based on Bacillus subtilis QST 713 or Bacillus amyloliquefaciens FZB42 (Borriss, 2015). These biocontrol products have an antagonistic effect towards unwanted microbes due to their secretion of antimicrobials, such as surfactin, fengycin or iturin (Ongena et al., 2005;Ongena and Jacques, 2008;Cawoy et al., 2014Cawoy et al., , 2015Saravanakumar et al., 2016). However, their precise mechanisms of action in fields are still unknown. Few studies have focused on the determinants of effective bioprotection. The surface colonization step and biofilm formation by biocontrol agents are highlighted in the publications cited in Table 1. These reports demonstrate that many biocontrol agents can form biofilms on crops and in the rhizosphere. It has also been shown that biofilm formation by biopesticides can be stimulated by plant root exudates (Espinosa-Urgel et al., 2002;Timmusk et al., 2005;Haggag and Timmusk, 2008;Khezri et al., 2011;Chen et al., 2013;Sang and Kim, 2014;Zhang et al., 2015), or by exposure of the microorganisms to antimicrobial products or stress (Bais et al., 2004;Selin et al., 2010;Fan et al., 2011;Xu et al., 2014;Chi et al., 2015;Wu et al., 2015;Zhou et al., 2016), but only a few studies have focused on biocontrol mechanisms that may be related to the properties of the mature biofilm itself, rather than the secretion of antimicrobials. Bacillus are ubiquitous spore forming bacteria predominantly found in soil. They are frequently used as biocontrol agents because they can sporulate and be stored for long periods (Branda et al., 2004;Borriss, 2015). Bacillus amyloliquefaciens FZB42 forms biofilms with little spatial organization in vitro (Fig. 1B), but exhibits a strong swarming capacity allowing a rapid surface colonization. For example, this strain can form biofilms on the fruiting body of Agaricus bisporus by forming bacterial clusters surrounded by extracellular matrix in contact with the mycelium of the carpophore ( Fig. 2A and B), as well as cell bundles (Fig. 2C). Fan et al. (2011) reported the induction of biofilm formation of B. amyloliquefaciens FZB42 by root exudates of maize and surfactin. Similarly, surfactin triggers biofilm formation by B. subtilis UMAF6614 on the melon phylloplane (Zeriouh et al., 2014). Root exudate of cucumber also drives the chemotaxis of Bacillus amyloliquefaciens SQR9 and induces the production of bacillomycin D that triggers biofilm formation in the rhizosphere (Xu et al., 2014). Similarly, stem lesions of rice induce the production of GltB, leading to the production of bacillomycin L and surfactin, both involved in the biofilm formation of B. subtilis Bs916 (Zhou et al., 2016).
Other biocontrol agents, such as endophytes, can also form biofilms. For example, some bacteria of the genus Paenibacillus form biofilms in wheat seeds and protect them from the invasion of Fusarium graminearum (D ıaz Herrera et al., 2016). Another family of biocontrol agents consists of the Gram-negative Pseudomonas, ubiquitous bacteria found in many plant rhizospheres (Table 1) (Espinosa-Urgel et al., 2002;Steddom et al., 2002;Matilla et al., 2010;Selin et al., 2010). Biofilm formation by Pseudomonas putida 06909 on citrus roots is induced by exudates of the phytopathogen Phytophthora parasitica. The bacteria colonize the mycelium of the fungi by feeding on its exudates and then form a protective biofilm on the citrus roots, which prevents new growth of the pathogen (Steddom et al., 2002;Ahn et al., 2007).
Living in a biofilm profoundly alters microbial properties relative to the planktonic mode of life (Whiteley et al., 2001;Shemesh et al., 2007;Vlamakis et al., 2008Vlamakis et al., , 2013Bridier et al., 2011b). Ongoing research is currently deciphering the molecular mechanisms involved in biofilm formation and their repercussions on biocontrol efficacy.

Molecular mechanisms involved in biofilm formation of biocontrol agents
Until recently, few studies in the biocontrol field have considered that the preferred lifestyle of microorganisms in the environment is the biofilm mode of life. The main features associated with biofilm formation are a diversification of cell types and increased tolerance to the fluctuation of environmental factors, boosting microbial persistence in the environment (Vlamakis et al., 2008(Vlamakis et al., , 2013Flemming et al., 2016). Bacteria and fungi can form biofilms on crops (as illustrated by the cultivated mushroom microbiota in Fig. 2), and in both cases, biofilm formation is composed of five major steps described in Fig. 1A (Costerton et al., 1999;O'Toole et al., 2000;Bianciotto et al., 2001;Davies, 2003;Triveni et al., 2012;Vlamakis et al., 2013;Pu et al., 2014;Haagensen et al., 2015;Li et al., 2015;Gulati and Nobile, 2016;Sheppard and Howell, 2016). Bacillus subtilis is the most highly documented bacterial model currently used to study the regulatory molecular mechanisms that govern biofilm formation. One specificity of the biofilm mode of life is the diversification of cell types. The presence of several bacterial subpopulations within the biofilm of B. subtilis has been clearly demonstrated, suggesting the spatiotemporal regulation of gene expression within such 3D structures (Vlamakis et al., 2008(Vlamakis et al., , 2013. Matrixproducing cells, surfactin-producing cells, flagellated motile cells and sporulated cells coexist in the same community ( Fig. 1A) and are spatially and temporally organized, differentially expressing specific sets of genes (Vlamakis et al., 2008;van Gestel et al., 2015;Mielich-S€ uss and Lopez, 2015;Wang et al., 2016). Indeed, the combination of surfactin-and matrix-producing cells enables the organization of cells into bundles (Fig. 2C). These interfacial microbial cables allow bacteria to visit surrounding spaces to increase the biofilm surface area for nutrient and oxygen intake (van Gestel et al., 2015). Several genes involved in the phenotypic heterogeneity have been identified and extensively analysed in this species. For example, hag, encoding a flagellar protein and expressed by a subpopulation of motile cells; tasA, eps, blsA expressed by matrix-producing cells; sfrA, involved in the production of surfactin lipopeptide; sigF, involved in cell sporulation; swr, involved in swarming motility; and the com genes, involved in genetic competence (Kearns et al., 2004;Verhamme et al., 2007;L opez and Kolter, 2010;Vlamakis et al., 2013;van Gestel et al., 2015;Mielich-S€ uss and Lopez, 2015). All these genes are directly or indirectly regulated by various regulators (e.g. Spo0A, DegU, ComA, SinI, SinR, AbrB), which can thus play a role in the regulation of plant bioprotection by B. subtilis (L opez and Kolter, 2010;Vlamakis et al., 2013;Cairns et al., 2014;Mielich-S€ uss and Lopez, 2015;Romero et al., 2016). Indeed, a mutation in a gene coding for a positive regulator (e.g. SinI) will decrease plant colonization and protection by diminishing attachment of cells to the roots, while mutations in a gene coding for a repressor (e.g. SinR, AbrB) will increase plant protection by an increased numbers of root-attached cells and the formation of hyper-robust biofilms (Chen et al., 2013). Major components of biofilm structure that ensure its cohesion are the extracellular polymeric substances (EPS) that are mostly composed of water and extracellular biopolymers (polysaccharides, proteins, DNA, lipids) (Flemming and Wingender, 2010). Many microbial EPS have a backbone composed of various biomolecules forming gels with various cohesive and viscoelastic properties. Trapping a high amount of water is important for microbial survival against desiccation on plant surfaces (Abdian and Zorreguieta, 2016). This organic slime also protects their inhabitants from the action of environmental pollutants and toxic compounds (Sutherland, 2001;Sheppard and Howell, 2016). Another important component of the biofilm structure are amyloid fibres formed by the protein TasA. These filaments bind cells together, leading to formation of complex structures in biofilms that can hold and concentrate molecules (e.g. quorum sensing signalling molecules), and may also form aggregates to defend cells within the biofilm (de Jong et al., 2009;Romero et al., 2010;Flemming et al., 2016).
Several studies have recently highlighted various physiological behaviours of Bacillus within biofilm communities, demonstrating the high level of complexity of their interactions (Mitri et al., 2011;Houry et al., 2012;Liu et al., 2015;Prindle et al., 2015;Flemming et al., 2016). Prindle et al. (2015) described a new function for ion channels in biofilms in which they conduct electrical signals via spatial propagation of potassium waves which depolarize adjoining cells and coordinate the state of the exterior and interior cells of the biofilm. In addition, Liu et al. (2015) discovered a 'collective oscillation' phenomenon involved in toxic chemical tolerance, based on metabolic codependency between exterior and interior cells of the biofilm, and consisting of cyclic pauses during biofilm growth which increase the availability of nutrients in the deepest layers. Houry et al. (2012) also demonstrated that motile bacilli, expressing a bactericide, can kill a heterologous biofilm population and then occupy the newly created space (Houry et al., 2012). Altogether, these cellular traits show the complexity of living associated with a surface in a spatially organized microbial community. They also give an overview of the protection that biofilms can provide to their inhabitants on plant surfaces. Those basic insights into biofilm development and interaction might pave our way towards various applications in the field of crop protection.

Spatial and nutritive competition
The spatial organization of biocontrol agent biofilms on crop surfaces varies depending on their genetic potential and the environmental conditions. For example, the biofilms of Bacillus biocontrol agents display wide architectural diversity between strains. In vitro, biofilms of B. amyloliquefaciens SQR9 and B. subtilis QST 713 exhibit the classical thick and highly organized 3D structure of bacilli ( Fig. 1C and D). In contrast, B. amyloliquefaciens FZB42 forms only thin structures of a few cell layers (Fig. 1B). However, this strain outcompetes the other two due to its swarming activity, leading to rapid coverage of the entire surface. This ability to rapidly colonize a niche (Fig. 3) has been described previously as a potential biocontrol mechanism and could be called upon for the strain B. amyloliquefaciens FZB42 (Timmusk et al., 2005;Haggag and Timmusk, 2008;Fan et al., 2011;Abd El Daim et al., 2015). For Paenibacillus polymyxa, root exudates of plants induce invasive root colonization and biofilm formation that invades sites that could be potentially occupied by pathogens, thus preventing them from settling onto the surface by forming a protective biofilm (Timmusk et al., 2005;Haggag and Timmusk, 2008;Abd El Daim et al., 2015). In an organized, 3D community, nutrients may be consumed faster than they can diffuse throughout the matrix (Breugelmans et al., 2008;Stewart and Franklin, 2008). Growth and survival in such a dense community is frequently associated with spatial competition. Habimana et al. (2011) explained the inhibition of Listeria monocytogenes by Lactococcus lactis on surfaces by considering the 3D race between the two species. Using a simplified individual-based model approach, they demonstrated that the differences in the growth parameters (lag phase and growth rate) of the two species could explain the observed inhibition of the pathogenic cells. Lactococcus lactis cells rapidly formed layers on the mixed community and completely saturated the interface in contact with the nutrient, limiting nutrient access to the pathogen. This example illustrates that part of a biofilm population can be starved within the bulk of the biostructure, even in a very rich environment. In addition, Liu et al. (2016) underlined that the specific interactions between species, such as strong or weak cooperation, exploitation or competition, contribute mostly to the spatial organization of biofilms, as these interactions create fitness effects in multispecies biofilms. Taking spatial organization and interspecies interactions within multispecies biofilms into account could increase our understanding of the interactions that take place in agrosystems that use biocontrol agents.

Antibiosis
The production of secondary metabolites by selected organisms is one of the best described mechanisms of agricultural microbial biocontrol (Ongena et al., 2005;Ongena and Jacques, 2008;Khezri et al., 2011;Cawoy et al., 2014Cawoy et al., , 2015Chen et al., 2015;Aleti et al., 2016;Raza et al., 2016;Saravanakumar et al., 2016). Bacillus genomes contain many genes involved in the production of secondary metabolites, recently compiled in an exhaustive classification of known and putative antimicrobial compounds (Zhao and Kuipers, 2016). Indeed, 4-5% of the genome of B. subtilis is allocated to the production of antibiotics and 8.5% of the genome of B. amyloliquefaciens FZB42 is allocated to the production of secondary metabolites with antimicrobial properties (Stein, 2005;Chen et al., 2009;Zhao and Kuipers, 2016). Many exhibit interesting antibacterial properties (e.g. difficidin), antifungal properties (e.g. bacillomycin D, fengycin and surfactin), or both (e.g. bacilysin) (Ongena and Jacques, 2008;Chen et al., 2009;Guo et al., 2014;Guo et al., 2015;Chowdhury et al., 2015;Luo et al., 2015;Kr€ ober et al., 2016). Most of the studies that have analysed the profile of antimicrobial production have relied on experiments using planktonic laboratory cultures. However, in B. amyloliquefaciens FZB42, the genes involved in bacilysin synthesis are overexpressed in biofilms, suggesting that the bacteria have a stronger antagonistic effect in their sessile mode of life (Fig. 3) (Kr€ ober et al., 2016). Similarly, in B. subtilis, the regulator NtdR controls the expression of the ntdABC operon, encoding enzymes involved in the biosynthesis of the antibiotic kanosamine (Inaoka et al., 2004;Vetter et al., 2013). A global transcriptomic study that compared gene expression of B. subtilis in various modes of life showed that this operon is strongly overexpressed in biofilms (Nicolas et al., 2012), suggesting the possible involvement of kanosamine in interspecies interactions in plurimicrobial biofilms. Volatile organic compounds (VOCs) can also trigger antimicrobial activity (Khezri et al., 2011;Audrain et al., 2015;Raza et al., 2016). Raza et al. (2016) demonstrated that VOCs of B. amyloliquefaciens SQR9 inhibited the growth of Ralstonia solanacearum on agar medium or in soil. Altogether, these studies show that secondary metabolites with antimicrobial activity can be overproduced (or simply produced) in the biofilm lifestyle, improving antagonistic biocontrol activity. The presence of EPS or amyloid fibres in biofilms can also locally concentrate these molecules and prevent their dilution into the ambient aqueous environment, and thus presumably increase the virulence of biocontrol agents against pathogens in agrosystems (Bianciotto et al., 2001;Romero et al., 2010;Xu et al., 2014;Flemming et al., 2016). Previous studies highlighted effects of antimicrobials secreted by one producer on crop protection. Santhanam et al. (2015) have also shown that in certain cases, a consortium of different antimicrobial producers is required for optimal plant bioprotection.

Microbial communication
Biofilms are dense, spatially organized communities of microorganisms with extensive forms of social life. They can use specific signalling molecules (autoinducers) that allow them to sense and communicate with the local surrounding populations (Fuqua et al., 1994). This quorum sensing (QS) is involved in various biological processes, such as swarming, stress tolerance (pH, antimicrobials, etc.), the production of secondary metabolites, horizontal gene transfer, colonization, biofilm maturation and the synthesis of virulence factors (Fuqua et al., 1994;Von Bodman et al., 2003). These signalling pathways are widely used in bacteria-bacteria and bacteria-eukaryote associations to regulate and coordinate their interactions. For example, N-acylhomoserine lactones (AHL) in Gramnegative bacteria, oligopeptides in Gram-positive bacteria and gamma-butyrolactones in species of the genus Streptomyces are autoinducers (Danhorn and Fuqua, 2007). In Pseudomonas aeruginosa, QS controls the expression of many bacterial functions. The LasI-LasR QS system, with the autoinducer synthase LasI and the signal receptor LasR, is involved in biofilm maturation and the organization of its 3D structure. A lasI mutant can initiate biofilm formation but is unable to form a mature biofilm, suggesting that LasI is involved in the late stages of biofilm development (Davies et al., 1998;De Kievit et al., 2001). Kavanaugh and Horswill (2016) demonstrated that the Staphylococci QS system, agr, is involved in biofilm disruption and dispersal.
In the field of biocontrol, it was shown that the protective activity of Pseudomonas fluorescens 2P24 on wheat was mainly controlled by the PcoI-PcoR QS system that governs biofilm formation, and not directly by the production of antimicrobial metabolites (Wei and Zhang, 2006). Such social behaviour has been shown to also govern intermicrobial and interkingdom interactions, such as communication interference represented in Fig. 3 or cooperative communication (Zhang and Dong, 2004;Kalia, 2013). For example, P. aeruginosa secretes 3-oxo-C12-HSL that affects the growth of C. albicans hyphae and inhibits its biofilm formation (Hogan et al., 2004). Amyloid fibres of the matrix form aggregates that can act as QS inhibitors by binding QS signalling molecules, and thus locally concentrate these molecules that can reach a 1000-fold higher concentration in the matrix than in planktonic cell environments (Charlton et al., 2000;Hense et al., 2007;Romero et al., 2010;Flemming et al., 2016). Other types of molecules can quench or degrade QS signalling molecules of another species (Zhang and Dong, 2004). Indeed, AHL-lactonase of Bacillus thuringiensis hinders the accumulation of AHL of Erwinia carotovora, thus decreasing the virulence of this bacterium on potatoes (Dong et al., 2004). In Bacillus, the lipopeptide surfactin, in addition to its antibiotic properties, can act like a signalling molecule to promote biofilm formation of the other relative Bacillus (L opez et al., 2009;Aleti et al., 2016). Volatile organic compounds emitted by prokaryotic and eukaryotic microbes are also part of their communication repertoire and can trigger global reprogramming of gene expression of their perceivers (Farag et al., 2013;Audrain et al., 2015;Raza et al., 2016). For example, acetic acid emitted by B. subtilis biofilms can promote biofilm formation of other physically separated B. subtilis cells and thus act as an important pathway of cell-cell communication (Audrain et al., 2015;Chen et al., 2015). These communication phenomena specific to biofilms, or amplified in biofilms, could be used to improve biocontrol in agrosystems.
Stress tolerance: adaptability properties and matrix as a protective shield Bacteria in biofilms exhibit specific physiologies associated with increased tolerance/resistance of the overall community to harsh conditions (Costerton et al., 1999;Whiteley et al., 2001;Shemesh et al., 2007;Bridier et al., 2011b). Physiological differences between sessile and planktonic cells are mostly related to differential patterns of gene expression associated with the two modes of life (Whiteley et al., 2001;Shemesh et al., 2007). Transcriptomic analysis of Streptococcus mutans, a bacterium associated with tooth decay, showed that 12% of the genome was differentially expressed in biofilm communities relative to their single-cell homologues. The differentially expressed genes coding for known functions are involved in transport, signalling, metabolism, protein and antimicrobial synthesis (Shemesh et al., 2007). Mark et al. (2005) evaluated the influence of exudates of two varieties of sugar beets on the transcriptomic profile of Pseudomonas aeruginosa PAO1. They showed that the expression of 516 genes was altered in response to one exudates and 451 to the other, and 134 genes responded to both. They found that genes coding for the synthesis of alginate, a major component of the biofilm matrix, were upregulated. These results suggest that P. aeruginosa PAO1 forms a biofilm in response to sugar beet exudates. They also showed that the transcriptomic profile of Pseudomonas aeruginosa PAO1 in response to exudates is variety dependent. Similarly, Matilla et al. (2007) compared the transcriptome profiles of Pseudomonas putida KT2440 in the planktonic exponential growth phase, the planktonic stationary growth phase, the sessile form, in sand microcosms and in the rhizosphere. They showed that transcriptomic profile of the planktonic mode of life in the stationary growth phase was the most different from that of the rhizosphere, whereas that of the biofilm lifestyle was more comparable. Indeed, they found that the gene involved in the synthesis of alginate was upregulated in the rhizosphere (Matilla et al., 2007; : additional data file). These 'omics' studies confirm the presence of biofilm formation in the rhizosphere or in response to plant exudates. These techniques should be increasingly considered in the study of microbial interactions in agrosystems and extended to metagenomics, metaproteomics and metatranscriptomic approaches as successfully performed in other fields (Blotti ere et al., 2013;Kaul et al., 2016).
Other cellular variations can occur during biofilm formation. In the early stages, Pseudomonas aeruginosa shows enhanced genetic diversification. Resulting phenotypes vary from cells involved in accelerated biofilm formation to those with enhanced dispersion properties. In the first case, biofilms exhibited pronounced spatial differentiation leading to rough and wrinkled colonies on agar. In the second case, the biofilms showed little spatial organization resulting in small and flat colonies (Boles et al., 2004). These differences illustrate the genetic plasticity of cells within a biofilm that enables them to cope with harsh environmental conditions. Stewart and Franklin (2008) also reported the existence of nutrient and oxygen gradients within biofilms creating a stratification of local microenvironments associated with a diversification of cell physiologies (Fig. 3). Population heterogeneity can generate multiple regulatory pathways leading, for example, to the phenomenon of competence in a subpopulation of cells, which coupled with the spatial proximity, facilitates horizontal gene transfer between biofilm cells. This can include the acquisition of plasmids carrying antimicrobial resistance genes (Witte, 2000;Abraham, 2011;Kung and Almeida, 2014;Liu et al., 2016). This diversification of cell types in biofilms strongly suggests that the biofilm lifestyle of biocontrol agents enables them to better adapt to, and resist, the hostile conditions encountered in agrosystems (the so-called insurance effects in Boles et al., 2004) than their planktonic counterparts.
Biofilms are ubiquitous and subject to harsh conditions, such as the presence of antimicrobials or desiccation. Stewart (2015) recently performed a meta-analysis of the literature from which he proposed that biofilm tolerance to antimicrobials depends neither on the size or chemistry of the antimicrobials nor the composition of the microbial biofilm or the material to which it adheres (Stewart, 2015). Based on his analysis, biofilm-associated tolerance is primarily related to the nature and composition of the biofilm matrix. Indeed, the composition of the matrix creates various meshes within the biofilm leading to a diffusion-reaction limitation that can reduce antimicrobial penetration and local biodisponibility (Fig. 3) (Stewart et al., 2001;Stewart and Franklin, 2008;Flemming and Wingender, 2010;  2011b; Stewart, 2015). Stewart (2015) also stressed that the presence of the matrix and the associated 3D organization renders slow-growth populations less sensitive to certain stresses than their planktonic counterparts. The matrix also plays a central role in interspecies and interkingdom interactions. Staphylococcus aureus and Candida albicans are often associated in human diseases, where they form a polymicrobial biofilm (Harriott and Noverr, 2009;Lindsay and Hogan, 2014). This association allows S. aureus to resist vancomycin, an antibiotic that is usually efficient against the planktonic form of S. aureus. The biofilm of C. albicans serves as the backbone of S. aureus microcolonies that form on their surface with the matrix of C. albicans covering and protecting the cells of S. aureus from the action of the antibiotic (Harriott and Noverr, 2009). Other reports have described the protection of Staphylococcus epidermidis by C. albicans (Adam et al., 2002) or of S. aureus by B. subtilis (Sanchez-Vizuete et al., 2015). In the latter case, the authors identified a single gene of the B. subtilis NDmed whose disruption suppressed the protective effect. This gene (ypqP) is likely involved in the production of matrix exopolysaccharides (Sanchez-Vizuete et al., 2015). Non-polysaccharidic components of the matrix can also contribute to the matrix shield; the amphiphilic protein BlsA produced by B. subtilis prevents the penetration of biocides by forming a hydrophobic 'raincoat' layer at the biofilm-air interface (Epstein et al., 2011;Kobayashi and Iwano, 2012). Biocontrol agents likely benefit from this protective shield on crop surfaces to avoid invasion by aggressive detrimental flora and protect crops.

Perspectives for sustainable agroecological approaches
Biocontrol mechanisms triggered by biological control agents in agriculture are not yet well understood, and even unknown in certain cases. A single biocontrol agent can use a combination of various biocontrol processes, best described for the strain B. amyloliquefaciens FZB42. The use of this bacilli can lead to antagonism, spatial and nutritional competition, antimicrobial production, the stimulation of plant growth and the induction of plant resistance (Timmusk et al., 2005;Haggag and Timmusk, 2008;Babalola, 2010;Fan et al., 2011;Xu et al., 2011;Kr€ ober et al., 2014;Chowdhury et al., 2015;Kr€ ober et al., 2016;Abd El Daim et al., 2015). The biofilm mode of life is still poorly taken into account in biocontrol, although it clearly plays a role in agrosystems and governs some of the observed beneficial effects. It would be informative, in the near future, to include phenotypic screening of the ability of strains to form biofilms as a rapid selection criterion of biocontrol agents. Several high-throughput screening tests that could be used for this application are described in the literature (Azeredo et al., 2016). Better genetic knowledge of the various cell functions in biofilms will also open doors to selection criteria based on the presence of specific genes involved in important and specific biocontrol functions (Kaul et al., 2016).
Invoking biofilm formation as a determinant of biocontrol efficacy could be a new attractive strategy to better control its beneficial effects. This could be achieved, for example, using natural biofilm promoting molecules that trigger the biocontrol agent QS response. In the case of B. subtilis, this could be surfactin, VOCs, specific microbial exopolysaccharides or crop extracts (Bais et al., 2004;Chen et al., 2013;Chen et al., 2015;Audrain et al., 2015;Zhou et al., 2016). This effect could also be obtained by adding a second strain (or more) with the ability to stimulate biofilm formation by the initial biocontrol agent, for example through VOC synthesis (Filoche et al., 2004;Audrain et al., 2015;Chen et al., 2015;Figueiredo et al., 2016). Attention should be paid, in this case, to select strains without antagonistic activity against each other or the beneficial action of the biocontrol agent, as previously reported for certain cocktails in the literature (Xu et al., 2011). The benefit of the biofilm mode of life for biocontrol agents could also be obtained using dedicated formulations, as suggested in other areas; for example, the development of new formulas grown as biofilms to orally administer probiotics (e.g. beads of agar or alginates) is under consideration (Rieu et al., 2014). Similarly, a system using bacteria-containing polymersomes, which permits rapid biofilm growth, has been developed for bioremediation to reduce the toxicity of environmental selenium contamination (Barlow et al., 2016).
Increasing evidence, based on available data from the agrosystem and biofilm fields, strongly suggests that the combination of features associated with the 3D biofilm mode of life should be considered when evaluating the performance of biocontrol organisms.